Periodic Reporting for period 1 - EPSKS (Efficient pore-scale kinetic simulation of gas flows in ultra-tight porous media)
Período documentado: 2018-07-15 hasta 2020-07-14
Gas flows in tight porous media with low-permeability are of great interest in numerous industrial applications, e.g. extraction of methane gas from ultra-tight rocks like shale, and geological storage of CO2 to tackle the global warming. Therefore, it is important to understand and quantify the gas flows in these low-permeability porous media with pore spaces as small as a few nanometers across.
In this project, novel mathematical methods are developed to directly study the gas flow in these unconventional porous media at the micro scale, for which the established classical models fail. These methods are based on the gas kinetic theory which deals with the distribution of gas molecules’ velocity, directly in the pore space of porous materials. We also develop advanced computer programs implementing the methods that can effectively run on massively parallel computer systems such as the UK’s academic supercomputer, AECHER. Using these tools, the classical models can be examined in details for their validity and new models can be developed based on the pore-scale studies.
In this project, novel mathematical methods are developed to directly study the gas flow in these unconventional porous media at the micro scale, for which the established classical models fail. These methods are based on the gas kinetic theory which deals with the distribution of gas molecules’ velocity, directly in the pore space of porous materials. We also develop advanced computer programs implementing the methods that can effectively run on massively parallel computer systems such as the UK’s academic supercomputer, AECHER. Using these tools, the classical models can be examined in details for their validity and new models can be developed based on the pore-scale studies.
The main work performed in this project includes
• Investigation of rarefied gas mixtures flows by considering quantum effects using ab initio calculation instead of using simplified molecular interaction models. Development of this Direct Simulation Monte Carlo method in the OpenFOAM. Through this study, it is found that the quantum effects can be significant for mixtures of low-weight molecules at low temperature.
• Develop a novel algorithm for accelerating the solving of non-linear gas-kinetic equation. The algorithm couples the solving of the gas-kinetic equation and macroscopic transport equation expressed as an equation system of averaged molecules’ properties such as density, velocity and temperature (the Navier-Stokes equation). The new algorithm can significantly reduce the computational cost of solving gas-kinetic equation.
• Investigated the local grid refinement technique in the context of pore-scale gas-kinetic simulation. It is found even though the local grid refinement technique can reduce the number of unknown variables to solve, but the efficiency of such approach cannot match with the gas-kinetic solver on the uniform Cartesian grid for simulation of complex geometry.
• Development and optimization of massively parallel simulation software, PIKS2D and PIKS3D.
• Development of a parallelization technique of the gas-kinetic equation on the graphics processing unit (GPU). The computer memory requirement of this algorithm is greatly reduced such that large-scale 3D simulation can be performed on a single GPU.
These results have been disseminated both as Journal publications (4 in total) and presentations/posters at academic conferences (3 in total). The computer codes developed has been released as open-source projects including the pore-scale solvers PIKS2D and PIKS3D (https://github.com/iPACT-Platform/PIKS2D(se abrirá en una nueva ventana) https://github.com/iPACT-Platform/PIKS3D(se abrirá en una nueva ventana)) and the ab initio DSMC collision procedure in OpenFOAM (https://github.com/zhulianhua/AbInitio(se abrirá en una nueva ventana)). The PIKS2D and PIKS3D codes are used extensively in the James Weir Fluid Lab for calculation the permeabilities of complex porous structures.
• Investigation of rarefied gas mixtures flows by considering quantum effects using ab initio calculation instead of using simplified molecular interaction models. Development of this Direct Simulation Monte Carlo method in the OpenFOAM. Through this study, it is found that the quantum effects can be significant for mixtures of low-weight molecules at low temperature.
• Develop a novel algorithm for accelerating the solving of non-linear gas-kinetic equation. The algorithm couples the solving of the gas-kinetic equation and macroscopic transport equation expressed as an equation system of averaged molecules’ properties such as density, velocity and temperature (the Navier-Stokes equation). The new algorithm can significantly reduce the computational cost of solving gas-kinetic equation.
• Investigated the local grid refinement technique in the context of pore-scale gas-kinetic simulation. It is found even though the local grid refinement technique can reduce the number of unknown variables to solve, but the efficiency of such approach cannot match with the gas-kinetic solver on the uniform Cartesian grid for simulation of complex geometry.
• Development and optimization of massively parallel simulation software, PIKS2D and PIKS3D.
• Development of a parallelization technique of the gas-kinetic equation on the graphics processing unit (GPU). The computer memory requirement of this algorithm is greatly reduced such that large-scale 3D simulation can be performed on a single GPU.
These results have been disseminated both as Journal publications (4 in total) and presentations/posters at academic conferences (3 in total). The computer codes developed has been released as open-source projects including the pore-scale solvers PIKS2D and PIKS3D (https://github.com/iPACT-Platform/PIKS2D(se abrirá en una nueva ventana) https://github.com/iPACT-Platform/PIKS3D(se abrirá en una nueva ventana)) and the ab initio DSMC collision procedure in OpenFOAM (https://github.com/zhulianhua/AbInitio(se abrirá en una nueva ventana)). The PIKS2D and PIKS3D codes are used extensively in the James Weir Fluid Lab for calculation the permeabilities of complex porous structures.
Progress has been made in both the developments of novel algorithms and high-performance computer codes. A novel numerical algorithm for the non-linear gas-kinetic equation has been designed which can efficiently get the converged solutions of both gas flow in both conventional and rarefied conditions. By using such an algorithm, the theoretical computing time of near-continuum flows can be reduced by several orders. A 2D and 3D high-performance computing codes have been developed for simulating pore-scale gas flows. The 3D code can scale up to 20,000 CPU cores in the UK’s national supercomputer, Archer. Using the novel tools, we have studied the pore-scale gas flows in digitally scanned samples ultra-tight sandstones. The deterministic approached in this project is much more efficient than existing high-fidelity tools such as the direct simulation Monte Carlo methods.
Towards the end of this project, the advanced numerical method will be published as a scientific paper in the leading journal of scientific computing, Journal of Computational Physics. The massively parallel codes will have been released as open-source projects on GitHub. The high-fidelity numerical data for real sandstone samples will be deposited in the Digital Rock Portal, a publicly accessible data portal for retrieval, sharing, organization and analysis of images of varied porous micro-structures.
The new tools and the high-fidelity results will advance the understanding gas transport in unconventional porous media, thus provide more accurate macroscopic models in large-scale simulations in different fields such as natural extraction, CO2 sequestration, and functional porous materials.
Towards the end of this project, the advanced numerical method will be published as a scientific paper in the leading journal of scientific computing, Journal of Computational Physics. The massively parallel codes will have been released as open-source projects on GitHub. The high-fidelity numerical data for real sandstone samples will be deposited in the Digital Rock Portal, a publicly accessible data portal for retrieval, sharing, organization and analysis of images of varied porous micro-structures.
The new tools and the high-fidelity results will advance the understanding gas transport in unconventional porous media, thus provide more accurate macroscopic models in large-scale simulations in different fields such as natural extraction, CO2 sequestration, and functional porous materials.